Chetan Naik

Stony Brook University

Master of Science (MS), Computer Science

January 1, 2014 – August 1, 2016

RV College Of Engineering

Bachelor of Engineering (BE), Electronics and Communications Engineering

January 1, 2007 – June 1, 2011

Microsoft

Principal Applied Scientist

June 1, 2026 – Present

Redmond, Washington, United States · Hybrid

Amazon

Senior Applied Scientist

July 1, 2021 – May 1, 2026

Greater Seattle Area

Amazon

Applied Scientist

September 1, 2016 – June 1, 2021

Greater Seattle Area

Natural Language Processing Lab, Stony Brook University

Research Assistant

January 1, 2015 – August 1, 2016

United States

Google

Analyst, Ad Traffic Quality Operations

July 1, 2011 – May 1, 2014

Greater Hyderabad Area

Word vectors for 4th grade processes using Skip-gram

December 1, 2015 – January 1, 2016

Implemented a Torch based system to train word vectors for 4th grade vocabulary. This system trains word vectors using skip-gram (uses word2vec to initialize vectors).

The Metis Challenge: Naive Bees Classifier - Driven Data.

September 1, 2015 – Present

Designed and implemented a CNN architecture using Theano to determine the genus of the bee based on photographs of the insects. Achieved an accuracy of 81.9% with just 25 iterations of training on CPU. World Rank: 28

Semantic Role-based Process Knowledge Acquisition.

September 1, 2015 – January 1, 2016

Designed and developed a method to acquire process knowledge using collective role inference across sentences. Semantic and lexico-syntactic features are used for Semantic Role Classifier. Collective role inference is performed using ILP with the help of role classifier and textual similarity scores. The proposed method acquires high quality process knowledge (F1 of 0.72 points). [To be submitted to ACL 2016 for publication.]

tMood

April 1, 2015 – Present

This project is developed during Bitcamp Hackathon. A Pebble app that analyses the sentiment of people around you using twitter feeds at your location. Backend for this project is developed in Python and hosted in pythonanywhere.com with Frontend developed in JavaScript.

Neera

April 1, 2015 – May 1, 2015

Neera is Rubiks Cube Solver Robot, designed and built using LEGO Mindstorms NXT and programmed using NXC language.

Question Answering System

February 1, 2015 – May 1, 2015

The task for this project involves answering 4th grade process recognition questions by frame alignment. The system takes a 4th grade process recognition question with answer choices as input and produces the correct answer along with the confidence as the output. - Baseline Solution: Finding word overlap (similar to BOW) between question and answer to get word overlap score and ranking the answers based on this score. This answered 48.02% of the questions(152 questions) - Role Based Aligner: First, question and answer sentences are processed using a Semantic Role Labeller to get semantic roles. Then the roles are aligned by measuring the entailment between the entities in the roles to form a alignment score, which is later used to rank answers. This answered 62.50% of the answers. - Jacana Word Aligner: Question and answer sentences are sent directly to Jacana Aligner which employs semi-markov CRF to align the sentences and produce alignment along with a score. This is then used to rank answers. This answered 44.07% of the questions. (The low accuracy of this method is due to the data used in training the CRF. The domain of training data was different than the domain of the questions and answers.)

Predicting Yelp rating stars from review text

August 1, 2014 – December 1, 2014

The goal of this project was to predict reviews' star ratings on Yelp from review text. We built the following models that perform text analysis on review data to predict the rating stars. 1. Baseline Model: The most common rating, 3 stars, is the rating predicted by this model for all the reviews. 2. Term Frequency Model: In this model we use frequency of word occurrence to predict the review rating. 3. LDA + Sentiment Model: This model predicts rating using Latent Dirichlet Allocation (LDA) with an added sentiment layer by extracting topics and sentiment associated with the review from review text. 4. NMF + Sentiment Model: In this model, we predict review rating using Non-negative Matrix Factorization (NMF) with an added sentiment layer by extracting topics and sentiment associated with the review from review text. We achieved an accuracy of 61% in predicting review rating stars.

Predicting facial beauty using Convolutional Neural Network (CNN)

August 1, 2014 – December 1, 2014

In this project, we designed and trained a Convolutional Neural Network (CNN) to predict facial beauty. We configured a deep learning system based on Caffe (a fast framework for deep learning) and trained it to classify faces into one of the 5 classes based on facial beauty. The trained model predicts facial beauty with an accuracy of 50.32%.

Predicting the Super Bowl and College Football Champions of 2015

August 1, 2014 – December 1, 2014

The goal of this project is to build machine learning models to predict the winners of 2015 Super Bowl and the College Football Championship using historical data. We have predicted the outcome of football matches entirely using the knowledge of previous game statistics. We have used three different models to do this: 1. Baseline model : "Point Score Difference Model". In this model we use the score difference to predict winners of future games. 2. Linear Regression Model: In this model, we use linear regression to predict the point difference for each game. 3. PageRank Model: Here, we model the game data as a graph with nodes as teams and edges as score differences between the teams. We then use PageRank on this game graph to rank all the teams. This ranking is used to predict winners of future games.

Development of Controlling and Stabilizing Algorithms for a Quadcopter using LPC2148 ARM Controller

January 1, 2011 – June 1, 2011

The aim of this project was to build a low cost prototype quad-rotor and come up with system stabilization algorithm for the stabilizing the quad rotor. - Came up with a mathematical model of a quad-rotor to correct pitch and roll errors by changing motor thrust. - Used complementary filters to filter noisy sensor data from the inertial measurement unit and get meaningful information. - Was awarded the first place at R.V. College of Engineering for electronics undergraduate research project in May, 2011. - Based on this work, presented a paper entitled “Design of Quadcopter with closed loop control system” at Visvesvaraya Technological University, Belgaum, in 2011.

Project Vyoma

August 1, 2010 – June 1, 2011

Project Vyoma is a research effort at R. V. College of Engineering to design unmanned aerial vehicles. - Devised and built a tachometer based on a rotary encoder which could measure the speed of an IC engine up to 300 RPM. - Designed and developed the DaQ system which captured data during flight from various sensors like GPS, pressure sensor and inertial measurement unit into a SD card to be retrieved later for analysis by aerodynamics team. - Achieved 8th position among 70 teams from all over the world in the 'SAE Aero Design' event held in Atlanta, in 2011.